FeedExploreAlertsSavedProfile

Categories

AICybersecurityInfrastructureDatabaseTech Updates

Tech news that matters.

FeedExploreAlertsSavedProfile
Back to feed
AI·CriticalBreaking

Google Tests a New Standard for AI on Websites

A developer sits at a desk with two screens, one showing lines of code and the other showing a web browser.
Google logo
Google

TL;DR: Google is testing a new web standard, WebMCP, that allows AI agents to interact with websites directly. This creates a reliable way for AI to perform tasks, replacing older, error-prone methods like screen scraping.

By Neeraj Dhiman·2h ago·2 min read·updated 6m ago
Source

Key facts

Category
AI
Impact
Critical
Published
2h ago
Source
InfoQ

Full summary

Google is testing a new web standard that lets AI agents reliably interact with websites instead of using brittle screen scraping.

Google has introduced a proposed web standard called WebMCP, which is now available for testing in Chrome's Origin Trials. This new protocol is designed to create a standardized way for AI agents running in a browser to interact with websites. It allows a website to explicitly offer “tools,” such as specific JavaScript functions or HTML forms, directly to an AI. This means an agent can perform actions like filling out a form or clicking a button in a structured, reliable manner, as intended by the site's developers. This marks a significant change from current methods, where AI agents often rely on guesswork. Techniques like scraping a site's underlying code or reading pixels from the screen are often slow, expensive, and can easily break whenever a website's design is updated.

This development is important for anyone building or managing web technologies. For developers and founders, WebMCP could enable a new wave of AI products that can reliably automate complex online tasks, like booking travel or managing online accounts. For CTOs and security teams, the standard introduces a formal, predictable channel for AI interactions. While this could offer better control and monitoring compared to the chaos of screen scraping, it also creates a new API that must be properly secured. This proposal represents a fundamental shift from an AI guessing how to use a website to the website explicitly telling the AI how to use it, which has major implications for both functionality and security.

As an Origin Trial, WebMCP is still experimental and will evolve based on feedback from the developer community. Its long-term success will depend on broad adoption by both website creators, who must implement the standard, and the developers of AI agents. This proposal is a key step toward a web that is not just human-readable but also machine-operable in a standardized way. The industry will be watching to see if other browser makers like Mozilla and Apple show interest, as cross-browser support is crucial for any new web standard to become truly universal.

Why it matters

WebMCP could enable a new class of reliable AI assistants that automate web tasks, but it also introduces new development and security considerations for websites.

Business impact

This standard could create opportunities for new AI-powered services that automate complex online processes, while requiring businesses to adapt their websites to be 'AI-friendly' to remain competitive.

Tags

#AI#automation#google#chrome#web standards

Related on Notifire

  • ResearchAI fact-checking for generated content
  • Researchllms.txt
  • ResearchKubernetes security
  • ResearchSoftware supply-chain security

Primary source: InfoQ

Part of our research on

  • AI agents and agentic workflows →
  • Model Context Protocol (MCP) →

Tech intelligence for engineering teams

Short, verified briefings on AI, cybersecurity, infrastructure, and data — with the analysis and action steps that matter. Every briefing is sourced, fact-checked, and bylined to a named editor.

[email protected]Story tips & corrections welcomeHow we report →

The Notifire briefing

Verified tech intelligence in your inbox — AI, security, infra, and data.

The day's most important tech briefings. No spam, unsubscribe anytime.

Sections

  • AI
  • Cybersecurity
  • Infrastructure
  • Database
  • Tech Updates
  • Web3 & Chains

Newsroom

  • About Notifire
  • Editorial team
  • Editorial standards
  • Methodology
  • AI disclosure
  • Corrections

Resources

  • Explore
  • Research hubs
  • Comparisons
  • Tech glossary
  • FAQ
  • Alerts & watchlists

Follow

  • RSS feed
© 2026 NotifirePrivacyTermsCorrections
An independent, AI-assisted publication. Built at </Alpheric>
IntelligenceLive panel
Live

Top trending

Last 24h

    Popular tags

    Add to watchlist

    +OpenAI+Claude+PostgreSQL+Kubernetes+Cloudflare+AWS+CVE Critical

    Notifire score

    0–100 priority signal — combines impact, freshness, trending velocity, and source credibility.

  1. Atom feed
  2. LinkedIn
  3. X / Twitter
  4. Facebook
  5. Instagram
  6. YouTube